Research And Application Of Clustering Technology Based On Graph Theory

Clustering technology is an important research field in the direction of artificial intelligence and pattern recognition.Research and application based on clustering technology has long been used in all directions of society.The purpose of clustering technology is to gather the sample points of unknown tags in the data scene into corresponding categories according to the specific clustering methods and the internal relationship between the data.Graph theory is a basic model in mathematics.It is composed of basic elements such as points and edges.The clustering technique which based on graph theory takes points and edges as the starting point,points represent data elements,and edges describe various relationships between data elements.Spectral clustering method is a research branch of clustering technology.Based on graph theory,it builds a?generalized?feature system using similarity matrix between data samples,and then obtains data after clustering the feature vectors decomposed by the feature system.Spectral clustering essentially divides the relationship between data and data and finds the optimal division between data in graph theory.It is based on graph theory and has a global optimal solution at the same time.It has been applied in many fields.Based on recent related research results combine the development of mature theories,the following two main research tasks were carried out.?1?When the target data is affected by noise or pollution,the clustering algorithm tends to shift the desired clustering effect.In order to address such challenge,based on the classic spectral clustering algorithm,and by using the strategy of transfer learning,the transfer spectral clustering algorithm based on inter-domain F-norm regularization TSC-IDFR is proposed in this paper.For the data in the target domain,TSC-IDFR selects the referenced examples,of which the sampling size is the same as the data size in the target domain,from the source domain?historical data?by means of the principle of the K^th nearest neighbor.Then,in terms of the mechanism of inter-domain F-norm regularization,the matrix composed of the spectral eigenvectors of the selected referenced examples from the source domain is used to assist the spectral clustering on the target data.As such,TSC-IDFR successfully resolves the clustering on the target data set?target domain?even if it contains much noise.?2?Based on spectral clustering and semi-supervised learning,aiming at the influence of outliers,interference colors,and time consumption of large-sized and medium-sized image segmentation on the target data set,a spectral clustering framework with adjustable affinity and structural common constraints was proposed,and apply this framework to actual image segmentation experiments.This method can make full use of semi-supervised information and integrate prior knowledge into standardized spectral clustering.On the one hand,this framework is based on a specific sampling method and classical KNN algorithm,which greatly shortens the segmentation time of medium and large images.On the other hand,due to clever framework specifications and changes in balance factors,it is more flexible to adapt to any semi-supervised constraints scenes.